removed custom atomic op

cuda/atomics.cuh

-#pragma once
-
-#if defined(__CUDA_ARCH__) && (__CUDA_ARCH__ < 600 || CUDA_VERSION < 8000)
-static inline __device__ void atomicAdd(double *address, double val) {
-  unsigned long long int *address_as_ull = (unsigned long long int *)address;
-  unsigned long long int old = *address_as_ull;
-  unsigned long long int assumed;
-
-  do {
-    assumed = old;
-    old = atomicCAS(address_as_ull, assumed,
-                    __double_as_longlong(val + __longlong_as_double(assumed)));
-  } while (assumed != old);
-}
-#endif

cuda/weighting_kernel.cu

@@ -2,8 +2,6 @@
 #include <ATen/cuda/detail/IndexUtils.cuh>
 #include <ATen/cuda/detail/TensorInfo.cuh>
 
-#include "atomics.cuh"
-
 #define THREADS 1024
 #define BLOCKS(N) (N + THREADS - 1) / THREADS